MathDial: 2,861 Teacher-Student Tutoring Dialogues on Math Word Problems (ETH Zurich)

Provider
Macina, Daheim, Chowdhury, Sinha, Kapur, Gurevych, Sachan (ETH Zurich, TU Darmstadt)
License
CC-BY-4.0 (source)
Size
4 files
Format
JSON Lines (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/eth-nlped/mathdial)
Access
Free and public. No account with the data provider is needed.
Category
Mathematics
Source
https://arxiv.org/abs/2305.14536

About this dataset

MathDial is a dataset of one-to-one tutoring conversations grounded in grade-school math word problems from GSM8K. It was collected by pairing human teachers with a large language model that played a 7th-grade student holding a realistic misconception. There are 2,861 dialogues. Each records the problem text and correct answer, the student's incorrect solution and student profile, and the teacher's free-text description of the student's confusion. The teachers' own ratings are included too: whether the student solved the problem by the end (or the teacher had to reveal the answer), and how typical the confusion and interactions were on a 1 to 5 scale. Education researchers and learning scientists use it to study pedagogy in dialogue, for example when tutors scaffold and when they give away answers, and to train and evaluate tutoring models. Load it with Hugging Face datasets. MathDial is licensed CC BY 4.0, and the GSM8K problems are MIT. Cite Macina et al., MathDial, Findings of EMNLP 2023.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("eth-nlped/mathdial")
print(ds)

Data structure

File types in sampled files: JSON Lines

Files

Example file: test.jsonl

Analyze this dataset without downloading it

On Scigantic, MathDial: 2,861 Teacher-Student Tutoring Dialogues on Math Word Problems (ETH Zurich) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Mathematics datasets